Error: "Could not initialize plugin 'libnvidia-vgx.so' for vGPU" when powering on a virtual machine with an NVIDIA L40S card on ESXi
search cancel

Error: "Could not initialize plugin 'libnvidia-vgx.so' for vGPU" when powering on a virtual machine with an NVIDIA L40S card on ESXi

book

Article ID: 316307

calendar_today

Updated On:

Products

VMware vSphere ESXi

Issue/Introduction

  • Virtual machine fails to power on when an NVIDIA L40S vGPU device (nvidia_l40s-24q) is attached on an ESXi 8.0.2 host.

  • The virtual machine successfully powers on if the vGPU device is detached.

  • The failure is specific to NVIDIA L40S cards; NVIDIA L40 cards function correctly without issue.

  • Troubleshooting steps such as enabling SR-IOV in the BIOS, configuring the vgpu-shared parameter, and restarting the xorg service do not resolve the failure.

  • The memmapMaxRAMMB parameter is already configured to 1032192 as per the NVIDIA vGPU Software Release Notes for VMware vSphere, however, the issue persists.

  • The vmware.log file for the affected virtual machine records the following errors:

[YYYY-MM-DDTHH:MM:SS] In(05)+ vmx - Power on failure messages: Could not initialize plugin 'libnvidia-vgx.so' for vGPU 'nvidia_l40s-24q'.
[YYYY-MM-DDTHH:MM:SS] In(05)+ vmx - Module 'PCIPluginLate' power on failed.
[YYYY-MM-DDTHH:MM:SS] In(05)+ vmx - Failed to start the virtual machine.
[YYYY-MM-DDTHH:MM:SS] In(05) vmx - Vix: [mainDispatch.c:4210]: VMAutomation_ReportPowerOpFinished: statevar=0, newAppState=1870, success=1 additionalError=0.

Environment

VMware vSphere ESXi 8.0.x

Cause

The issue is caused by an NVIDIA display mode configuration conflict with the nvidia_l40s-24q vGPU profile on the ESXi host.

Resolution

The physical display mode on the vGPU must be disabled using the NVIDIA Display Mode Selector Tool following the steps below.

  1. The virtual machine must be powered on without attaching the vGPU device. This ensures the system boots without initializing the GPU display capabilities.

  2. The NVIDIA Display Mode Selector Tool must be installed on the guest operating system. The tool can be obtained from the official NVIDIA documentation or driver package.

  3. The following command must be executed within the NVIDIA Display Mode Selector Tool to disable the physical display mode on the vGPU:

    displaymodeselector --gpumode physical_display_disabled

  4. The virtual machine must be rebooted. A system restart is required for the changes to take effect.

  5. The virtual machine must be powered off, and the vGPU device must be attached.

Additional Information

The following articles can be referenced for similar known issues with NVIDIA vGPU:

The VMware Compatibility Guide (VCG) and VMware vSphere Release Notes must be reviewed prior to deployment to verify hardware compatibility and identify any known issues with specific GPU architectures.

Powered by Wolken Software